Re: [AMBER] running RISM/MD on multi processors

From: Tyler Luchko (Lists) <"Tyler>
Date: Tue, 2 Aug 2016 14:46:02 -0700

Hi Joy,

This is a limitation of sander, not 3D-RISM. There is no easy way to get around this.

You *may* not benefit from 96 processors anyway. The parallelization depends on FFTW, which does a slab decomposition along the z-axis. Parallelization will not be linear in general but if you have fewer than 4 grid points per process, there will be a large slow down. 3D-RISM should report to you the size grid it is using.


> On Aug 2, 2016, at 2:26 PM, Jingyi Yan <> wrote:
> Hi all,
> I am running MD simulations coupled with 3d-RISM calculations. When I run
> the system on 96 processors, it reported that: Must have more residues than
> processors!
> Since RISM/MD simulation is super time consuming, and I am wondering if I
> can use more processors to calculate.
> Thanks!
> Joy
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Tue Aug 02 2016 - 15:00:02 PDT
Custom Search